Race, Income, and College in 25 Years:
Evaluating Justice O’Connor’s Conjecture
Alan Krueger, and Jesse Rothstein, Princeton University and NBER, and
Sarah Turner, University of Virginia and NBER
In Grutter v. Bollinger, Justice O’Connor conjectured that in 25 years affirmative
action in college admissions will be unnecessary. We project the test score distribution of black and white college applicants 25 years from now, focusing on the role
of black–white family income gaps. Economic progress alone is unlikely to narrow
the achievement gap enough in 25 years to produce today’s racial diversity levels
with race-blind admissions. A return to the rapid black–white test score convergence of the 1980s could plausibly cause black representation to approach current
levels at moderately selective schools, but not at the most selective schools.
Even in the absence of continuing bias, the legacies of de jure segregation and racial discrimination in the United States create gaps in income
and educational attainment between blacks and whites that will persist for
at least the near future. Because there are comparatively few black highschool graduates with exceptional achievement test scores, the most selective colleges would admit very few black students if admission criteria
We thank Bill Bowen for suggesting this topic, and Martin Kurzweil and
Nirupama S. Rao for help with the Mellon Expanded College & Beyond data.
The UCLA Center on Education Policy and Evaluation, the Russell Sage Foundation, the Carnegie Corporation, and the Princeton Industrial Relations Section
provided generous research support. We are grateful to Orley Ashenfelter, Martin
Kurzweil, Gary Solon, and Jacob Vigdor for excellent comments on an earlier
version of this draft.
Send correspondence to: Jesse Rothstein, Princeton University, Wallace Hall,
Princeton, NJ 08544; E-mail: [email protected].
American Law and Economics Review
doi:10.1093/aler/ahl004
Advance Access publication July 25, 2006
ª The Author 2006. Published by Oxford University Press on behalf of the American Law and Economics
Association. All rights reserved. For permissions, please e-mail: [email protected]
282
Race, Income, and College in 25 Years
283
relied exclusively on measures of academic achievement currently in use.
Many colleges and universities attempt to offset the gaps in the credentials
of black and white applicants by giving an advantage in admission to black
applicants over white applicants with similar academic records.
The rate at which racial gaps in pre-collegiate academic achievement
can plausibly be expected to erode going forward is a matter of considerable uncertainty. Justice O’Connor, in her opinion in the Grutter v. Bollinger
case (which held affirmative action in college admissions to be a constitutional means of attaining diversity in student populations under some
circumstances), takes a firm stand on this question: ‘‘We expect that 25
years from now, the use of racial preferences will no longer be necessary to
further the interest approved today.’’ Our goal in this article is to evaluate
the plausibility of Justice O’Connor’s conjecture.
We project the elite college applicant pool 25 years from now, under
assumptions—which we believe are optimistic—about the rate at which
existing racial gaps in economic circumstances and pre-collegiate educational achievement will close. Our analysis focuses on two important
margins: Changes in the relative income distributions of black and
white families and narrowing of the test score gap between black and
white students with similar family incomes. Progress on each margin
can be expected to reduce the racial gap in qualifications among
students pursuing admission to the most selective colleges. We use
existing estimates of the speed of regression toward the mean income
across generations to project the future black–white family income gap,
and extrapolate from past trends in test score convergence between
black and white students to project the conditional-on-income test
score gap. Combining the two projections with a unique national
dataset on SAT test takers, we predict the distribution of black test
scores relative to those of whites in future years.
After projecting the pool of likely applicants, we simulate the effects
of alternative admission policies—including the current race-conscious
system, race-neutral SAT-based admissions, and class-based affirmative
action—on the racial composition of admitted students. We neglect
other aspects of the college ‘‘pipeline:’’ We consider application rates
only as a parameter in our projections, and we make no effort to
predict matriculation decisions. We also restrict our attention to
black and white students, as immigration and assimilation issues
284
American Law and Economics Review V8 N2 2006 (282–311)
make Hispanic and Asian applicant pools substantially more difficult
to define and forecast. In light of the U.S.’s distinct historical legacy of
racial policies, the representation of African Americans in elite colleges
is of unique interest. Finally, we restrict our analysis to selective
institutions, as race-conscious admission policies are only relevant
where admissions are competitive. Thus, we say little about aggregate
college attainment trends, which primarily reflect outcomes at nonselective, open access institutions.
Our results represent a baseline expectation in the event that current
trends continue, and are obviously sensitive to policy shifts that change
relative outcomes of African American and white students. Changes
in class size, in school effectiveness, and in income inequality, for
example, would all have important effects on black representation
among potential applicants, admissions, and matriculants at selective
colleges.
The legacy of racial inequality in academic and economic opportunity forms the background of the admissions debate and of this analysis. The first section of this article outlines the racial gap in outcomes
at the elementary and secondary levels. The second section discusses
the role of affirmative action policies in college admissions. The third
section describes our methods and data and presents our forecasts of
test score distributions and admissions outcomes 25 years from now.
1. Racial Inequality in Pre-Collegiate Achievement
In 1970, the average 17-year-old black student scored more than one
standard deviation below the average white student on the first National
Assessment of Education Progress (NAEP) assessment (Figure 1). Progress since then has been slow and episodic, and essentially stopped around
1990. Trends at the 90th percentile of the black and white distributions,
perhaps more relevant for admission to selective colleges, are similar to
those at the mean. Today, the black–white gap stands at about three
quarters of a standard deviation in reading, and even higher in math.
An obvious partial explanation for the persistence of this gap is the
continuing gap in economic resources between black and white students’
families. Black workers earn substantially less, on average, than do whites,
though the differential has slowly narrowed. The average black male
Race, Income, and College in 25 Years
285
1.4
Standard deviation units
1.2
1
0.8
0.6
0.4
Math, Difference in means
Math , Difference in 90th percentile
Reading, Difference in means
0.2
0
1968
Reading, Difference in 90th percentile
1973
1978
1983
1988
1993
1998
Year
Figure 1. Trends in Black–White gaps in Student Achievement, National
Assessment of Education Progress (NAEP) Test, Age 17.
Source: National Center for Education Statistics, 1999.
earned 38% less than the average white male in 1960, 25% less in 1980, and
26% less in 2000.1
Improved labor market performance for adult black workers has been
slow to translate to improved economic circumstances for black children, as
deterioration in black family structures has partially offset increases in
individual earnings. In 2002, 35% of black children resided in families with
two parents, down from 59% in 1968. The trend was less dramatic for white
children, 74% of whom resided in families with two parents in 2002, down
from 89% in 1968.2 As a result, the gap in total family income between black
and white children has hardly moved in three decades: The ratio of the
median income for black families with one child to the median income for
similar white families was 0.63 in 1967 and 0.62 in 2001.3
1. Card and Krueger (1992) and authors’ calculations from 2000 Census data.
Juhn, Murphy, and Pierce (1991) and Card and Krueger (1993) argue that continuing skill convergence during the 1980s was offset by increasing inequality.
2. http://www.census.gov/population/socdemo/hh-fam/tabCH-3.xls, Tables CH-2
and CH3
3. http://www.census.gov/hhes/income/histinc/incfamdet.html, Tables F9A and
F9B
286
American Law and Economics Review V8 N2 2006 (282–311)
The racial gap in test scores among families with similar incomes also
remains sizable. Phillips and coauthors find that only about two thirds of
the black–white gap in young children’s scores is explained by even a rich
set of family background measures (Phillips et al., 1998). The source of the
black–white gap within income groups is not well understood. For the
purpose of our primary projections, we treat the within-income test score
gap as a black box and consider the likely trends in this unexplained gap.
We do, however, explore one potential partial explanation, that black
students attend inferior elementary and secondary schools.
2. Affirmative Action and College Access for Black Students
Many colleges began instituting affirmative action policies, which give
preferences in admissions to minority applicants, in the late 1960s. These
policies have been the subject of periodic judicial scrutiny since their
inception. In 2003, the Supreme Court handed down decisions in two
cases (Gratz v. Bollinger and Grutter v. Bollinger) challenging the consideration of race in University of Michigan admissions. These decisions
upheld the use of some forms of preferences, justified both by the oncampus benefits derived from a racially diverse educational environment
and by the societal importance of producing a cadre of minority leaders
and professionals (see Goldstein, 2004, for further analysis). The court was
hesitant to endorse the use of racial preferences in perpetuity, however,
and Justice O’Connor predicted in her Grutter opinion that the need for
such preferences would disappear within 25 years.
Proactive efforts to recruit and admit students from underrepresented
groups have produced sizable gains in the representation of black students at
the most selective colleges and universities (Bowen and Bok, 1998). Blackwell
(1987) finds that black representation at Ivy League universities increased
from 2.3% in 1967 to 6.3% in 1976; shares in other ‘‘prestigious’’ institutions
increased as well, from 1.7% to 4.8%. Affirmative action preferences could not,
however, have had large effects on overall college enrollment rates: The vast
majority of colleges and universities in the U.S. do not employ the sorts of
selective admission policies that are a precondition for the use of preferences
(Kane, 1998).
The precise mechanics of admission policies at selective colleges are
not widely understood. The courts have prohibited mechanical
Race, Income, and College in 25 Years
287
approaches to affirmative action—quotas (Reagente of the University
of California v.Bakke) or ‘‘points’’ (Gratz)—and most selective colleges
employ difficult-to-quantify ‘‘holistic’’ evaluations. Still, by examining
average admission probabilities among groups defined by important
determinants, like SAT scores, it is possible to get an idea of the roles
of race and academic qualifications in admissions. Bowen, Kurzweil,
and Tobin (2005) have generously provided us extracts from the data
used in their study of the role of academic and nonacademic factors in
admissions at a number of selective institutions. We present results for
four groups of colleges and universities: most selective, highly selective, and moderately selective private institutions, and elite public
universities.4 It must be emphasized that these labels are relative
characterizations; even the least selective group in our typology is
extremely selective by any national standard.
Admission profiles are shown in Figure 2. Broadly, the difference
between the likelihood of admission of black and white applicants
with the same SAT score is largest in the middle of the applicant
pool. At the very top—SAT scores over 1500—both black and white
applicants are very likely to gain admission, though racial preferences
remain substantial at the most selective colleges; at the bottom, few
students of either race are admitted. Adding additional ‘‘controls’’
would not alter the basic pattern of differences in admission by race.
Without the preferences indicated by Figure 2, there would be many
fewer black students admitted to the Expanded College and Beyond
(ECB) institutions. The policy question at the center of our empirical
analysis is whether there will be enough black students among the
highest scorers 25 years from now to yield a critical mass of minority
students under race-blind admission criteria. In effect, if the same
admission profiles are applied to black and white applicants, then
4. These data are from the Expanded College and Beyond (ECB) study
assembled by the Andrew W. Mellon Foundation, and describe the 1995 admissions cycle. The most selective private institutions are Harvard, Princeton, and
Yale; the highly selective are Columbia, the University of Pennsylvania,
Swarthmore, and Williams; and the moderately selective are Barnard, Bowdoin,
Middlebury, Oberlin, Pomona, and Wellesley. Public universities are Pennsylvania
State University and the University of Virginia. Confidentiality requirements prevent a
more disaggregated presentation.
Admits / Applications
1
1200
1300
1200
1300
1500
Admission Rate, White
1400
1600
1600
0.4
0.5
0.6
0.7
0.8
0.9
1
0
800
0
800
0.1
SAT Bin Beginning
1400
1500
Admission Rate, Black
0
800
0.1
0.2
1100
SAT Bin Beginning
1100
0.1
0.2
0.3
0.3
1000
1000
Admission Rate, Black
Admission Rate, White
0.4
0.5
0.6
0.7
0.8
0.2
900
Selective 2
900
Public Universities
1
0.9
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0
800
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Figure 2. Admission Rates by Type of Institution and Race.
Source: Authors’ Calculations from Expanded College and Beyond.
Admits / Applications
Admits / Applications
Admits / Applications
900
Selective 3
900
1000
1000
1200
1300
1200
1300
SAT Bin Beginning
1100
SAT Bin Beginning
1100
Admission Rate, Black
Admission Rate, White
Selective 1 Universities
1500
1400
1500
Admission Rate, Black
Admission Rate, White
1400
1600
1600
288
American Law and Economics Review V8 N2 2006 (282–311)
Race, Income, and College in 25 Years
289
what will be the representation of black students in the pool of students likely to receive admission offers in 25 years?
3. Projections
3.1. Methods and Data
Our approach in projecting the pool of applicants to selective colleges
and universities takes into consideration expected changes in the relative
distributions of black and white family incomes and in the pre-collegiate
achievement of black and white youth over the next quarter century. We
begin by projecting the relative distributions of black and white family
income, which are likely to converge somewhat in coming years. As SATs
depend heavily on family income, increases in black families’ relative
incomes will lead to increases in black students’ relative scores. This will,
however, almost certainly understate the overall progress in black relative
test scores.
The black–white gap in test scores among children with the same family
incomes is large (Jencks and Phillips, 1998) but mutable. In one set of
estimates, we assume that the conditional (on income) gap will fall at the
same rate over the next 25 years as has the unconditional gap over the last
quarter century. This is almost certainly too optimistic, as the entire gain
over the last 25 years occurred in the first 10 years of that period; more
recent data indicate a growing black–white test score gap. We also present
somewhat less optimistic estimates based on simulations in which we
assume that the school quality gap between black and white students is
equalized.5 Again, reasonable people may differ in their projections of the
likely rate of future convergence in black–white achievement; our estimates
are meant to indicate what sort of assumptions one would need to make to
obtain desired admission results.
5. The school quality gap might be closed, for example, by perfectly integrating
the schools. Even in such an implausible scenario, our estimates overstate the
impact on black–white gaps: Income affects test scores in part by purchasing access
to better schools, so we double-count by combining income-based progress and
integration gains. Across-the-board increases in school quality are also possible.
Black students may be more sensitive to school quality than are whites (Krueger
and Whitmore, 2002), though we judge the prospects for substantial closing of the
black–white gap through this channel as limited.
290
American Law and Economics Review V8 N2 2006 (282–311)
Using data on current application behavior and admission probabilities
by student race and qualifying test score, we simulate admissions to the
four groups of selective colleges, with both current test score distributions
and our simulated distributions with alternative admission rules. These
simulations allow us to assess whether it is plausible, as Justice O’Connor
conjectured, that there will be enough high-scoring blacks in 25 years that
race-blind admission rules will produce as diverse a class as is admitted
using affirmative action today.
We rely on several data sources. For our estimate of the baseline distribution of college preparedness by race and family income, we use a
dataset containing observations on approximately one third of students
from the high school class of 2000 who took the SAT college entrance
exam. We use the public-use microdata sample of the 2000 Census (from
the Integrated Public Use Microdata Series—Ruggles et al., 2004) as our
source for current family income distributions and as the basis for our
projections of future distributions. Our estimates of the rates of change in
black–white test score gaps derive from the NAEP Long-Term Trend data
(NCES, 1999), a time series of scores on an unchanging test over the last
quarter century. Finally, application and admission outcomes for a set of
selective colleges and universities are calculated from the aforementioned
ECB dataset, which describes the cohort entering college in the fall of 1995;
we compute denominators for application rates by comparing these data to
data on 1995 SAT test takers.
3.2. Projections of Test Score Distributions
Figure 3 displays the distribution of SAT scores for blacks and whites in
2000. Although the distributions are similar in shape, the mean for black
students is about 200 points lower than that for white students, resulting in
extreme underrepresentation of black students at the highest scores. The
bottom panel of the figure shows the fraction of students at each SAT
score who are black. Blacks are 14.3% of (black and white) SAT takers but
are substantially underrepresented at every score above the grand mean of
1,000. This underrepresentation is most severe at the rightmost tail: In the
2000 cohort depicted here, there were about 250 white students who earned
perfect scores of 1,600 on the SAT but only two black students. Similarly,
there were nearly 22,000 white test takers with scores of 1,400 or above but
Race, Income, and College in 25 Years
291
Panel A
0.12
Whites
Blacks
Fraction SAT Test Takers
0.10
0.08
0.06
0.04
0.02
0.00
400
600
800
1000
1200
1400
1600
SAT Score
Panel B
0.8
0.7
0.6
Share Black
0.5
0.4
Overall black share of black and white
test takers (14.3%)
0.3
0.2
0.1
0
400
600
800
1000
1200
1400
1600
SAT Score
Figure 3. Distribution of SAT Scores, Black and White Test Takers.
Panel A: Dearity by race. Panel B: Bkck share. Source: Authors’ Calculations
from Test Takers Database, 2000 Cohort.
292
American Law and Economics Review V8 N2 2006 (282–311)
only 322 black test takers. To maintain the current representation of black
students at the most selective schools—where 65% of admitted students
have SAT scores above 1,400—under admission policies that do not consider race directly, black academic progress must substantially reduce the
underrepresentation in Figure 3.
3.2.1. Income Convergence. Among families with children aged 15–17
years in the 2000 Census, the average black family’s income is about
$34,500 less than the average white family’s, and the median black family
falls below the 25th percentile in the white family income distribution.6
This gap is attributable partly to differences in family structure between
races—57% of black families and 24% of white families containing a 15- to
17-year-old child have only one parent present—and partly to differences
in labor market participation and outcomes.
Estimates of the intergenerational transmission of incomes indicate
that, on average, somewhere between 40% and 60% of the gap between a
father’s (log) income and the mean (log) income will be closed by his son
(Mazumder, 2000, is at the low end of this range; Solon, 1999, is at the high
end). Estimates of the correlation of income or wealth across generations
at the family level are within the same range (Chadwick and Solon, 2002;
Solon, 2002). We do not separately model changes in family structure and
in incomes conditional on family structure, though the estimates cited here
for the intergenerational correlation of family incomes reflect both. To the
extent that black family structures converge toward white structures more
quickly than would be expected given current income gaps, we will tend to
understate the potential for family income convergence.
Although it is between-group convergence that is the focus of our
analysis, most current estimates of the intergenerational transmission of
income derive from samples of blacks and whites and assume no distinction between within-group and between-group parameters. There are several reasons to expect that this might be unreasonable. First, black incomes
did not converge toward those of whites for the first several centuries of
black presence in North America, although the correlation in income
between fathers and sons was undoubtedly well below 1. This indicates
6. We have also estimated our projections using as the relevant universe families
with one member between the ages of 36 and 50, with similar results.
Race, Income, and College in 25 Years
293
that mean regression can go on within groups even when—for reasons of
discrimination or otherwise—there is little or no convergence between
groups. Second, the rate of within-group convergence may well be different
for blacks than for whites.7 Hertz (2005) argues that persistent poverty
among very low income black families drags down estimates of intergenerational mean reversion, suggesting that pooled estimates may overstate
between-group convergence. Finally, even if we accept the accuracy of
current estimates for between-group convergence, there is no guarantee
that the rate of intergenerational transmission of income in the future will
be the same as it was in the past. For example, an increase in the return to
skill could cause intergenerational mean reversion to slow.
Nonetheless, we would not have gone too far astray had we used a .40–.60
coefficient of intergenerational mobility to project the black–white income
gap in the recent past. Consider the following: In 1969, the average 30- to 39year-old black male worker—who had attended separate and unequal
schools and entered the labor force before the Civil Rights Act of 1964
barred discrimination—earned 37% less than the average white worker.
Based on the coefficient of intergenerational transmission of earnings
alone, this gap would have been expected to close to 15%–22% for the next
generation. The actual earnings gap for men in their 30s in 1999—roughly
one generation beyond the same age range in 1969—was 19%, well within the
range of the forecast.8 This accurate forecast may just be a coincidence, but
at least it was not wildly off. Applying it to another generation, the black–
white earnings gap would be projected to close to 6%–13% when members of
the third generation reach their 30s, around a quarter century from now.
For our projections, we take the middle of the consensus range and assume
that the gap in mean log incomes between white and black families will shrink
by one half over the next generation. We believe that this is more likely to
overstate than to understate the rate of convergence.
Figure 4 displays histograms of black and white family incomes in 2000
from the decennial Census. The actual black–white gap in mean log incomes
was nearly 0.77 (corresponding to a gap of 54%, or about $13,000, at the mean)
in that year, so the above assumptions imply that it will shrink by about
7. Mazumder (2000) finds little indication of differences, though his estimates
are imprecise.
8. This example is from Krueger (2003).
294
American Law and Economics Review V8 N2 2006 (282–311)
0.38 log points over the next quarter century. Our projections of the future
black income distribution (relative to whites), then, are obtained by inflating the
income of each black family in the 2000 data by 32% (=ln(1 + 0.38)). The
resulting distribution is indicated by the dashed line on the figure.9 With this
projection, the fraction of black families with incomes between $80,000 and
$100,000 will increase by 69% (from 4.7% to 8.0%), whereas the fraction with
incomes between $25,000 and $30,000 will fall by 83% (from 7.6% to 6.3%).
These changes provide re-weighting factors that can be used to estimate the
effect of economic progress on black test score distributions.
Our SAT sample records students’ self-reports of their family
income in 13 categories, so we compute re-weighting factors from
0.12
0.10
White actual
Black actual
Black counterfactual
Relative Frequency
0.08
0.06
0.04
0.02
0.00
$0
$50,000
$100,000
$150,000
$200,000
$250,000
Family Income Category
Figure 4. Black and White Family Income Distributions, 2000 Census.
Source: Authors’ Calculations from 2000 Decennial Census Public Use
Microdata Sample (Ruggles et al., 2004).
9. Census demographic projections indicate that the population of 17-year-old
blacks will grow by 14.9% between 2000 and 2025, whereas the corresponding white
population will shrink by 8.4% (from http://www.census.gov/ipc/www/usinterimproj/).
This demographic growth, not incorporated in Figure 4, can be expected to expand the
number of blacks (and shrink the number of whites) at each income level. We do not
include this in our projections, as expansions of black population shares arguably have
commensurate effects on the ‘‘equal representation’’ goalposts.
Race, Income, and College in 25 Years
295
the census data for each of these categories to represent the projected
family income distribution. To illustrate, recall that we project that
with a stable white income distribution the number of black families
with incomes between $80,000 and $100,000 will rise by 69% over the
next quarter century. In our counterfactual SAT distribution, we
count each black SAT taker with a family income in this range 1.69
times. This has the expected result that re-weighted black average
scores are higher (by about 19 points) than are current averages, as
an increase in the fraction of blacks from high-income families produces an upward shift in the SAT distribution. It also implies a small
(about 0.7%) increase in the black SAT-taking rate, as SAT-taking
rates are higher among higher income families. The projected test
score distribution is shown as the ‘‘income counterfactual’’ series in
Figure 5.10
The way to interpret our projection is that we have held everything
constant—the distribution of white incomes and the distribution of test
scores by income and race—except that we increase black families’ incomes
by the amount predicted from intergenerational convergence. Of course, real
income growth will raise both groups’ incomes over the next quarter century. This will alter the figures on the x-axis of Figure 5 but, absent changes
in inequality, not the shape of the distribution. Our approach implicitly
indexes income growth by the mean white family’s income growth. Similarly, our models of test scores and admissions treat the white test score
distribution as stable. The distribution may well change, but we implicitly
assume that admission standards adjust to maintain current admission
probabilities of students at each percentile of the test score distribution.
3.2.2. Test Score Convergence. Our first set of estimates assume zero
convergence of test scores conditional on family income, which might be
thought of as a reasonable lower bound for test score prospects for the next
quarter century. An upper bound is provided by assuming that black–white
10. One disconnect between our simulation and the educational process is
worth noting. Although the reasons why family income affects student performance
on the SAT are unclear, it is quite likely that the entire stream of family income over
a child’s time at home is relevant, not just income in the year he or she takes the
SAT. Unfortunately, we lack data on family income in earlier years. Many of the
convergence estimates in the literature apply to permanent, not annual, income.
296
American Law and Economics Review V8 N2 2006 (282–311)
0.12
Baseline
Income counterfactual
Income + NAEP Progress Counterfactual
Fraction of Current Black SAT Takers
0.10
Income + Integration Counterfactual
0.08
0.06
0.04
0.02
0.00
400
600
800
1000
1200
1400
1600
SAT Score
Figure 5. Predicted Effects of Income Growth on the Distribution of Test
Scores of Blacks.
gaps close as much within-income groups in the next 25 years as did
unconditional gaps over the last 25 years. Referring back to Figure 1, in
the long run there has been some narrowing in the gap between black and
white test scores among 17-year-old students. Linear regression lines fit to
the age 17 NAEP black–white difference at the 90th percentile—this is likely
most informative about the SAT-taking population—indicate that blacks
have gained 0.44% of a SD annually relative to whites on the math exam and
1.59% of a SD annually on the reading exam.
The verbal and math components of the SAT exam have standard deviations of approximately 100 points each, so the NAEP trend, if it continues at
the same rate, would imply that the black–white gap in SAT scores should
close by just over 50 points over the next quarter century. To incorporate this
trend into our analysis, we simply add this many points to each black student’s
SAT score in the 2000 data, after re-weighting the data to reflect income
convergence. The dashed line in Figure 5 shows the resulting distribution.
This almost certainly overestimates the extent of black score growth over
the next quarter century. As Figure 1 indicates, essentially all of the progress
over the last 25 years in NAEP scores occurred in the 1980s, and the gap grew
Race, Income, and College in 25 Years
297
during the 1990s. It would take substantial optimism to assume that future
progress will occur at the rate seen over the full NAEP period rather than the
much slower rate seen recently, particularly as we are assuming that this
progress will be in addition to that generated by income convergence.
3.2.3. Equalizing School Quality. An alternative approach to projecting the
distribution of black scores is to imagine specific interventions into the
educational process. One particularly ambitious—and, it must be admitted,
wholly politically implausible—intervention might be to fully integrate
schools. The average black SAT taker attends a high school where 52% of
(black and white) SAT takers are black, whereas black students are only 8%
of SAT takers at the average white SAT taker’s school (Card and Rothstein,
2006). Schools are not just separate but also of unequal quality. It is unlikely
that interventions in the educational process can have larger effects on black–
white test score gaps than to close this school quality gap. Thus, projections
that assume the quality gap will be eliminated provide an alternative
optimistic view of the prospects for conditional-on-income progress.
We construct a crude estimate of quality as the school fixed effect in a
regression of SAT scores on a rich vector of student background characteristics.11 Estimated this way, the student-level SD of school quality is 69
SAT points, and the median black SAT taker attends a school whose
quality is 32 points lower than that of the median white SAT taker’s
school. To implement the ‘‘integration’’ approach, we match corresponding percentiles of the black, white, and overall school quality distributions
and re-assign the overall quality distribution to both blacks and whites.
That is, we assume that black students who currently attend the best
schools attended by black students will, in the integrated counterfactual,
attend the best schools overall, and the same for whites. Our re-assignment
has the effect of closing the black–white gap in mean scores (in the income–
re-weighted data) by 30 points, the gap in median scores by 34, and the gap
at the 90th percentile by 24.
11. The regression includes full interactions of individual gender, race, and 13
family income dummies and of race with 100 (10 mother’s by 10 father’s) parental
education dummies. Our approach ascribes both peer groups and any other schoollevel components of test score variation to school quality (Rothstein, forthcoming).
In particular, we overstate the importance of schools if there are any important
unobserved aspects of individual background that vary across schools.
298
American Law and Economics Review V8 N2 2006 (282–311)
0.16
0.14
Baseline
# Blacks / (# Black + # White)
0.12
Black share of test takers at
baseline
Income counterfactual
Income + NAEP progress counterfactual
0.1
Income + integration counterfactual
0.08
0.06
0.04
0.02
0
1000
1100
1200
1300
1400
1500
SAT Bin Beginning
Figure 6. Black Share of (Black and White) SAT Takers, by SAT Score.
Figure 6 shows the fraction of students at each score who are black in
each of the counterfactual simulations (presented in Figure 5): Income
growth only, income growth plus NAEP convergence, and income growth
plus integration. By construction, the first simulation has the smallest
effect, increasing the number of high-scoring (1,400 or above) blacks by
about 54% over its current low level. The integration scenario is next,
producing (in combination with income convergence) a 109% increase in
high-scoring blacks. The most optimistic scenario is the one using NAEP
trends, in which the number of high-scoring blacks increases by 225%.
Even under this counterfactual, however, the proportion of blacks scoring
above 1,400 will be about one quarter of the corresponding proportion of
whites, with more extreme underrepresentation at higher scores.
3.3. Admission Projections
Our interest is in how the projected changes in the relative distribution
of the academic achievement (measured by test scores) of black and white
students will alter the relative representation of black and white students
Race, Income, and College in 25 Years
299
among those likely to be admitted to selective colleges and universities
under race-blind admission policies. To address this, we must convert SAT
distributions to admission rates. The observed admission decisions of
colleges and universities provide admission profiles, by SAT, for composite
institutions of varying selectivity. We focus on four composite profiles of
admission outcomes defined as: most selective private (Selective 1:
Harvard, Princeton, and Yale), highly selective private (Selective 2: Columbia,
the University of Pennsylvania, Swarthmore, and Williams), moderately
selective private (Selective 3: Barnard, Bowdoin, Middlebury, Oberlin,
Pomona, and Wellesley), and selective public (University of Virginia and
Pennsylvania State). We model expected admission to each of these
‘‘composite’’ schools rather than to the individual institutions.
To calculate expected admissions under each of our simulations, we
simply multiply
X
ðNumber of test takersjr Application ratejr
Expected admissionsr ¼
j
Admission ratejr Þ
for each race (r) and SAT level (j). Under the current regime, both application
rates and admission rates differ by race. Under a race-neutral policy, blacks
and whites with the same test scores would face the same probability of
admission, conditional on application. We implement this by assigning the
admission profile observed for whites to blacks, in effect assuming that both
black and white students will face the admission probabilities indicated by the
solid lines in Figure 2.12 We also consider alternative policies that provide
admission advantages to students from the lowest income families (what some
have called ‘‘class based affirmative action’’ [CBAA] ).
It is important to emphasize that our calculations are inherently static,
as we do not explicitly model the changes in individual application behavior and college admission policies that a shift to race-neutral admissions
would entail. Most importantly, a large shift in admission probabilities
12. If application behavior is unchanged, the elimination of racial preferences
will reduce the total number of admittees. As the share of students admitted under
affirmative action is small, this effect is as well. Nevertheless, to the extent that
colleges lower the race-blind admission standards to compensate, we will very
slightly overestimate the effect of affirmative action on black admission shares.
300
American Law and Economics Review V8 N2 2006 (282–311)
would likely lead to responses in black students’ decisions about where to
apply. At each SAT score, black students currently are substantially more
likely than are whites to apply to the most selective institutions (see Figure 7,
using the institutional data). This disparity is smaller at less selective
institutions, where it largely disappears at the highest SAT scores.
Substantial increases in the rate at which high-scoring black students
apply to elite colleges are unlikely: As Figure 7 indicates, application rates
among these students are already quite high. Indeed, using the SAT data,
we calculate that well over half of blacks with SATs above 1,500 send their
scores—a proxy for application—to Harvard alone.13 This does not speak
well for the prospects for increasing minority representation through better
outreach to potential applicants, and we therefore focus on changes in the
number of high-scoring blacks as the primary potential source of black–
white convergence in the number of qualified applicants.
A plausible explanation for the existing racial differences in application
rates is that black students respond to their admission advantages at
selective colleges and universities. If this is indeed the explanation, then
one might expect application rates to converge as admission probabilities
do. One piece of evidence that weighs against this expectation, however, is
that at least in the short-run the elimination of race-conscious admission
policies in Texas and California appears not to have altered the pattern of
applications of high-achieving black students (Card and Krueger, 2005). In
any event, in addition to estimates based on current race-specific application rates, we also consider a scenario in which black application rates
come to resemble those of whites.
Table 1 presents simulations based on the assumption that black
application behavior remains as it is today. The first row summarizes
the actual representation of black students among those admitted at
the four institutional composites in 2000.14 We define representation
13. Card and Krueger (2005) found substantial similarity between patterns seen
in actual applications and those obtained by examining score reports.
14. Note that our analysis focuses on the pool of students admitted to composite institution types, not the actual representation of students in the entering
cohort. There are presently substantial differences between black and white students in matriculation, which we expect would change with policy shifts such as the
elimination of race-conscious admissions. We believe the most judicious strategy is
to avoid projections of enrollment which necessarily rely on parameters that are
difficult to project.
Applications / Test Takers
0.00
800
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.00
800
0.05
0.10
0.15
0.20
1000
2
1100
900
1000
1100
Predicted Application Rate, Black
Predicted Application Rate, White
Application Rate, Black
Application Rate,
White
Selective
900
Predicted Application Rate, Black
Predicted Application Rate, White
Application Rate, Black
Application Rate, White
1200
SAT Bin
1200
SAT Bin
1300
Selective 2
1300
1400
1400
PublicPUBLIC
1500
1500
1600
1600
0.00
800
0.05
0.10
0.15
0.20
0.25
0.30
0.00
800
0.20
0.40
0.60
0.80
1.00
1.20
Figure 7. Application Rates by Type of Institutions.
Source: Authors’ Calculations from Expanded College and Beyond.
Applications / Test Takers
Applications / Test Takers
Applications / Test Takers
0.25
900
900
1200
1000
1100
SAT Bin
1200
Predicted Application Rate, Black
Predicted Application Rate, White
Application Rate, Black
1300
1300
Selective 1
SAT Bin
Selective 3
1100
Application Rate, White
1000
Predicted Application Rate, Black
Predicted Application Rate, White
Application Rate, Black
Application Rate, White
Selective 1
1400
1400
1500
1500
1600
1600
Race, Income, and College in 25 Years
301
0.061
0.073
School quality convergence
Counterfactual income distribution,
school quality convergence
NAEP, National Assessment of Education Progress.
0.075
0.088
Observed income distribution, NAEP progress
Counterfactual income distribution, NAEP progress
0.125
0.308
0.336
0.545
0.064
0.086
0.089
0.118
0.069
0.051
0.000
0.169
0.053
0.064
0.171
0.118
Counterfactual income distribution
for families with children
Status quo admission rates
(with race preferences)
Observed income distribution
Projected race neutral admissions
(using current white admission rates)
Observed income distribution
Most Selective (S1)
Highly Selective (S2) Moderately Selective
(S3)
0.000
0.110
0.293
0.316
0.557
0.153
0.072
0.092
0.094
0.117
0.078
0.060
0.142
0.000
0.146
0.386
0.410
0.700
0.214
0.059
0.071
0.072
0.087
0.062
0.051
0.094
0.182
0.463
0.504
0.835
0.254
0.000
B/B+W Sh. of Gap B/B+W Sh. of Gap B/B+W Sh. of Gap B/B+W Sh. of Gap
Closed
Closed
Closed
Closed
Public (P)
Table 1. Expected Share of Black Students Relative to White Students, Alternative Income and Test Score Distributions,
Maintaining Current Black Application Patterns
302
American Law and Economics Review V8 N2 2006 (282–311)
Race, Income, and College in 25 Years
303
as the number of black students divided by the number of black
students plus white, non-Hispanic students. This omits students of
other races and ethnic groups. Thus, where we calculate that 16.1%
of black and white students admitted to Selective 1 institutions are
black, a more inclusive calculation would indicate that only 10.4% of
all students admitted to these schools are black (and 54.5% are white,
non-Hispanic).
The second row of the table summarizes the black share of admissions
under the ‘‘race neutral’’ counterfactual, in which the observed white
admission profile is applied to both blacks and whites. With current test
score distributions, this would reduce the representation of black students
by more than two thirds, from 17.1% to 5.1%, at the most selective private
institutions (Selective 1). Projected declines in the representation of black
students are by no means limited to the most selective institutions and are
estimated at 55% for public institutions, 58% for the highly selective
private institutions, and 46% for the moderately selective private institutions. These are the gaps which black relative academic progress must close
to realize Justice O’Connor’s prediction that race preferences will no
longer be necessary to accomplish what affirmative action is needed to
accomplish today.
Row 3 of Table 1 applies the same race-neutral admissions rule to the
first counterfactual SAT distribution, assuming income convergence but
no additional progress in test scores. This produces small gains in the
representation of black students relative to what would be seen today
with the same admission rules. For each institutional composite, we
show the share of the gap between current representation of black students
and that which would be seen with race-neutral admissions that our
projected income-driven convergence would close. Only about one fifth
of this gap is closed at the public, most selective and highly selective
composite institutions. Gains are slightly larger at the moderately selective
private institutions, closing a quarter of the gap. It appears that reasonable
income convergence will not, on its own, allow for the abolition of affirmative
action without severely affecting the representation of African American
students at elite colleges.
When we allow for progress through reductions in the black–white test
score gap, the expected representation of black students among those
admitted expands considerably. These shares are shown in rows 4 and 5 of
304
American Law and Economics Review V8 N2 2006 (282–311)
Table 1, first without and then with expected income convergence (which
may involve some double counting, as part of past test score convergence
may be attributable to income changes). For public and less selective institutions, narrowing of the test score gap combined with income convergence
would go a considerable distance toward reproducing today’s levels of
diversity, if it can be assumed that application behavior does not change.
This is less true at the most selective private institutions, where the black
share would remain substantially lower than is observed today.
The last two rows of the table present estimates of the representation of
black students relative to white students under the alternative counterfactual
of school integration (or of equalization in school quality). As one might
infer from Figures 5 and 6, these projections are between the estimates with
income convergence alone and those with both income and historical test
score convergence, closing about 30% of the gap created by the shift to raceneutral admissions. Given the extent of progress that the school integration
scenario entails, the degree to which it is surpassed by the test score convergence scenario underscores the optimism inherent in the latter.
Table 2 presents an identical analysis under the assumption that black
application rates come to resemble those seen today among whites with
similar SAT scores. Any declines in application rates of high-scoring black
students would exacerbate the drop in black representation produced by
moving to a race-neutral admission policy. Black shares are lower in each
simulation in this table, but the effects of income and test score convergence are similar.15
3.4. Alternative Admission Rules
Racial minorities are not the only underrepresented group in elite colleges.
Students from middle and lower income families, regardless of race, are also
less likely to have the SAT scores needed for admission under current
admission rules. Some observers (e.g., Kahlenberg, 1996) have proposed
that elite colleges implement CBAA, giving preferences to low-income students akin to those now given to racial minorities. As blacks (and Hispanics)
tend to have lower incomes than whites, some have even suggested that
15. If the elimination of affirmative action causes black application rates to
become like those of whites, then the appropriate comparison is of row 1 in Table 1
with rows 2 through 8 of Table 2.
0.045
0.055
School quality convergence
Counterfactual income distribution,
school quality convergence
0.081
0.205
0.222
0.366
0.022
0.031
0.033
0.046
0.036
0.095
0.106
0.194
0.046
0.000
Sh. of Gap
Closed
NAEP, National Assessment of Education Progress.
Note: Share of gap closed is computed relative to simulations using current race-specific application rates (row 1).
0.056
0.068
Observed income distribution, NAEP
Counterfactual income distribution, NAEP
0.023
0.016
0.112
0.038
0.047
0.057
0.083
B/B+W
0.171
0.000
Sh. of Gap
Closed
Most Selective (S1)
0.118
Counterfactual income distribution
for families with children
Status quo admissions rates
(with race preferences)
Observed income distribution
(current application rates)
Observed income distribution
(white application rates)
Projected race neutral admissions
(using current white admissions rates)
Observed income distribution
B/B+W
Public (P)
0.027
0.037
0.038
0.051
0.029
0.021
0.048
0.142
B/B+W
0.047
0.128
0.137
0.246
0.067
0.000
Sh. of Gap
Closed
Highly Selective (S2)
0.034
0.044
0.045
0.057
0.037
0.028
0.047
0.094
B/B+W
0.090
0.242
0.255
0.446
0.132
0.000
Sh. of Gap
Closed
Moderately Selective
(S3)
Table 2. Expected Share of Black Students Relative to White Students, Alternative Income and Test Score Distributions,
Assuming Blacks Adopt Current White Application Patterns
Race, Income, and College in 25 Years
305
306
American Law and Economics Review V8 N2 2006 (282–311)
income-sensitive policies could be a means of admitting more black students
without the legally tenuous consideration of race per se. It is not clear,
however, whether the race–income correlation is strong enough to make
family income a useful proxy for race. If it is not, then CBAA will be a
blunt tool for achieving racial diversity. What is more, it can only become
blunter in the future, as ongoing narrowing of black–white income gaps will
make black students even less identifiable in the income distribution and
further worsen income’s efficacy as a proxy for race.
To illustrate the potential effects of CBAA both today and in the
future, we apply this type of admission rules to our simulated SAT
distributions. Twenty-one percent of SAT takers report family incomes
of $35,000 or less roughly comparable with the 15% who are black. We
thus model a CBAA admission rule as giving the same admission
advantages to students with family incomes below $35,000 as is today
given to black students. The results are illustrated in Table 3.16 Consideration of family income does, indeed, increase the representation of
black students. However, because black students are only moderately
overrepresented among the additional students admitted under incomebased preferences, and because the students brought in under these
preferences comprise only a small share of the baseline class, the effect
on total black representation is relatively modest. Under our baseline
assumption, with no convergence in income or test scores, the black
share rises from 5.1% to 7.0% at the most selective private institutions
and from 5.3% to 7.2% at public institutions.
When we turn to the simulation with income convergence, black overrepresentation among low-income SAT taker shrinks, as does the effect of
CBAA on black admission shares. This is partially offset when we add to
the simulation test score convergence, which produces substantial increases
in the number of low-income blacks with mid-range scores, where
16. More low-income students than black students have SATs in the middle of
the distribution, where preferences have the largest effects (Figure 3). As a result, a
shift from race- to income-based admissions would, if the size of the preference is
held constant, lead to more total admissions. To the extent that colleges respond to
this by tightening admission standards across the board (the alternative would be to
expand the admissions pool, by as much as one quarter), black admission shares
will be lower than those shown in the tables, as more black than white admittees are
marginal admits.
5.1
15.3
7.0
6.0
21.2
7.7
5.1
20.9
6.2
5.3
20.5
7.2
773
4,199
5,266
686
5,952
5,764
438
6,201
14,817
2,091
16,908
% Black
3,426
Note: Simulations assume status quo application behavior.
Selective 1
No preferences
Income preferences (<$35k)
Additional admits
Total admits
Selective 2
No preferences
Income preferences (< $35k)
Additional admits
Total admits
Selective 3
No preferences
Income preferences (< $35k)
Additional admits
Total admits
Public
No preferences
Income preferences (<$35k)
Additional admits
Total admits
Number of
Admits
Baseline
2,009
17,001
14,992
421
6,252
5,831
660
6,026
5,366
752
4,246
3,494
Number of
Admits
17.2
7.7
6.4
17.7
6.9
6.2
18.0
8.9
7.8
13.0
8.0
6.9
% Black
Income Convergence
Table 3. Projections of Admission Pool with Income-Based Affirmative Action
2,149
17,544
15,395
462
6,453
5,991
740
6,349
5,609
840
4,526
3,686
Number of
Admits
22.6
10.5
8.8
25.0
9.8
8.7
26.9
13.5
11.7
22.1
13.7
11.8
% Black
Income + Test Score
Convergence
Race, Income, and College in 25 Years
307
308
American Law and Economics Review V8 N2 2006 (282–311)
preferences are strong. The share of black students among the additional
students admitted under an income-based policy rises to nearly 22.1% at
the most selective universities, 26.9% at schools in the next selectivity
band, 25% at moderately selective schools, and 22.6% at public universities. The CBAA beneficiary pool remains small, however, so we see
relatively modest changes in the representation of black students.
4. Discussion
Affirmative action is a significant feature of admission policies at
the nation’s most selective colleges today, but it may not be in the
future. The legacies of separate and unequal schooling and of labor
market discrimination are reflected in the academic preparation of the
current generation of black students. In an equal opportunity society,
the effects of past discrimination on current generations will eventually asymptote to zero, though there is substantial uncertainty
about the rate at which this might be expected to occur. In Grutter,
Justice O’Connor suggests that affirmative action in admissions can
only be considered constitutional as a temporary policy, and she
forecasts it will not be necessary 25 years hence.
To provide a quantitative assessment of Justice O’Connor’s speculation, we consider the racial composition of today’s admitted students
as a baseline and ask whether foreseeable progress in black economic
and educational success can plausibly be expected to permit race-blind
admission rules to reproduce today’s race-conscious results in a quarter
century.
We are most confident in predicting that economic progress alone will
not yield as much racial diversity as is generated with today’s race-sensitive
admission policies. Under plausible assumptions about changes in the
income distribution of black families in the next 25 years, the representation of black students at selective colleges under race-blind admissions will
be only 42% of the status quo. Put another way, black economic gains over
the next quarter century can be expected to provide only about 17% of the
incremental representation that is provided by affirmative action today.
This conclusion is not much changed if the future admissions policy is
assumed to incorporate CBAA, in which students from families with low
incomes are given preferences analogous to those given to racial minorities
Race, Income, and College in 25 Years
309
today. The correlation between race and family income, while strong, is
not strong enough to permit the latter to function as a useful proxy for race
in the pursuit of diversity. Moreover, the value of income as a proxy for
race can only decline with increases in black incomes.
Similarly, we judge it unlikely that universities will be able to compensate for the abolition of race-based preferences through increased outreach
toward and recruitment of minority students. Our exercise suggests that
there are simply be too few high-scoring black students and that they
already apply to the most selective colleges at rates far exceeding those of
white students with similar scores.
We do find some reason for optimism, however, in our projections
incorporating an extrapolation of past increases in black students’ test
scores relative to whites’. Our most hopeful simulations assume that,
conditional on income, black students’ scores relative to white students’ will rise at the same rate over the next quarter century as a
linear trend fit to the 1970–98 experience. As part of the past progress
in narrowing the test score gap may reflect income convergence, these
projections probably involve some double counting. Nevertheless, in
this scenario, and if black student application behavior is assumed
stable, we find that race-blind admission policies may approach the
black representation achieved by affirmative action, at least in some
categories of colleges. This projection is likely upwardly biased
because the last 25 years saw two distinct regimes, with rapidly closing
black–white gaps in the first period and deterioration in black relative
performance since 1990. To extrapolate a linear trend a full quarter
century into the future is to assume a dramatic turnaround from
recent patterns and sustained growth over a long period. On the
other hand, if we could somehow return to and sustain the rapid
rate of progress seen in the 1980s, then the future will be much
brighter than even our optimistic forecasts.
As an indication of the difficulty of achieving racial diversity on highly
selective college campuses without affirmative action, we consider the
effects of a wholly implausible intervention producing the complete integration of the nation’s secondary schools. This, we estimate, would produce only a small fraction of the test score gains that would be needed to
make Justice O’Connor’s prediction a reality. Clearly, substantial progress
in increasing black students’ pre-collegiate performance is critical to any
310
American Law and Economics Review V8 N2 2006 (282–311)
hope of eliminating the need for affirmative action within the next generation. Absent such progress, the elimination of racial preferences in admissions,
today or 25 years from now, will lead to substantial declines in black representation at the nation’s most selective colleges and universities. Our simulations, crude as they are, lead us to agree with Justice Ruth Bader Ginsburg’s
concurring opinion in Grutter, ‘‘From today’s vantage point, one may hope,
but not firmly forecast, that over the next generation’s span, progress toward
nondiscrimination and genuinely equal opportunity will make it safe to sunset
affirmative action.’’
References
Blackwell, James. 1987. Mainstreaming Outsiders: The Production of Black Professionals, 3rd ed. Dix Hills, NY: General Hall.
Bowen, William, and Derek Bok. 1998. The Shape of the River: Long Term Consequences of Considering Race in College and University Admission. Princeton,
NJ: Princeton University Press.
Bowen, William, Martin Kurzweil, and Eugene Tobin. 2005. Equity and Excellence
in Higher Education. Charlottesville: University of Virginia Press.
Card, David, and Alan Krueger. 1992. ‘‘School Quality and Black-White Relative
Earnings: A Direct Assessment,’’ 107(1) The Quarterly Journal of Economics
151–200.
Card, David, and Alan Krueger. 1993. ‘‘Trends in Black-White Earnings Revisited,’’ 83(2) American Economic Review 85–91. Papers and Proceedings of the
Hundred and Fifth Annual Meeting of the American Economic Association,
May.
Card, David, and Alan Krueger. 2005. ‘‘Would the Elimination of Affirmative
Action Affect Highly Qualified Minority Applicants? Evidence from California
and Texas,’’ 58(3) Industrial and Labor Relations Review 416–34.
Card, David, and Jesse Rothstein. 2006. ‘‘Racial Segregation and the Black-White
Test Score Gap.’’ National Bureau of Economic Research Working Paper
12078, Cambridge, MA.
Chadwick, Laura, and Gary Solon. 2002. ‘‘Intergenerational Income Mobility
Among Daughters,’’ 92(1) American Economic Review 335–44.
Goldstein, Joel K. 2004. ‘‘Beyond Bakke: Grutter-Gratz and the Promise of
Brown,’’ 48 Saint Louis University Law Journal 899–954.
Hertz, Tom. 2005. ‘‘Rags, Riches and Race: The Intergenerational Economic
Mobility of Black and White Families in the United States.’’ In S. Bowles,
H. Gintis, and M. Osborne, eds., Unequal Chances: Family Background and
Economic Success. Princeton, NJ: Russell Sage Foundation and Princeton
University Press, 165–191.
Race, Income, and College in 25 Years
311
Jencks, Christopher, and Meredith Phillips, eds. 1998. The Black-White Test Score
Gap. Washington, DC: The Brookings Institution.
Juhn, Chinhui, Kevin M. Murphy, and Brooks Pierce. 1991. ‘‘Accounting for the
Slowdown in Black-White Wage Convergence.’’ In Marvin Kosters, ed., Workers and Their Wages. Washington, DC: AEI Press, 107–43.
Kahlenberg, Richard D. 1996. The Remedy: Class, Race, and Affirmative Action.
New York: Basic Books.
Kane, Thomas J. 1998. ‘‘Misconceptions in the Debate over Affirmative Action in
College Admissions.’’ In Gary Orfield and Edward Miller, eds., Chilling Admissions: The Affirmative Action Crisis and the Search for Alternatives. Cambridge,
MA: Harvard Education Publishing Group, 17–31.
Krueger, Alan. 2003. ‘‘The Supreme Court Finds the ‘Mushball Middle’ on Affirmative Action,’’ New York Times, July 23.
Krueger, Alan, and Diane Whitmore. 2002. ‘‘Would Smaller Classes Help Close the
Black-White Achievement Gap?’’ In John Chubb and Tom Loveless, eds.,
Bridging the Achievement Gap. Washington, DC: Brookings Institute Press,
11–46.
Mazumder, Bhashkar. 2000. ‘‘Earnings Mobility in the US: A New Look at Intergenerational Mobility.’’ Mimeo, Federal Reserve Bank of Chicago.
National Center for Education Statistics (NCES). 1999. ‘‘Results Over Time—
NAEP 1999 Long-Term Trend Summary Data Tables,’’ http://nces.ed.gov/
nationsreportcard/tables/Ltt1999/
Phillips, Meredith, Jeanne Brooks-Gunn, Greg J. Duncan, Pamela Klebanov, and
Jonathan Crane. 1998. ‘‘Family Background, Parenting Practices, and the
Black-White Test Score Gap.’’ In Christopher Jencks and Meredith Phillips,
eds., The Black-White Test Score Gap. Washington, DC: Brookings Institution
Press, 103–145.
Ruggles, Steven, Matthew Sobek, Trent Alexander, Catherine A. Fitch, Ronald
Goeken, Patricia Kelly Hall, Miriam King, and Chad Ronnander. 2004. Integrated Public Use Microdata Series: Version 3.0 [Machine-Readable Database].
Minneapolis, MN: Minnesota Population Center, http://www.ipums.org
Rothstein, Jesse. Forthcoming. ‘‘Good Principals or Good Peers? Parental Valuation of School Characteristics, Tiebout Equilibrium, and the Incentive Effects of
Competition Among Jurisdictions.’’ National Bureau of Economic Research
Working Paper 10666, Cambridge, MA, American Economic Review.
Solon, Gary. 1999. ‘‘Intergenerational Mobility in the Labor Market.’’ In Handbook of Labor Economics, Vol. 3A. Amsterdam: Elsevier, 1761–800.
Solon, Gary. 2002. ‘‘Cross-Country Differences in Intergenerational Earnings
Mobility,’’ 16(3) Journal of Economic Perspectives 59–66.